Is arsenic a major determinant of health aid disbursement? An evaluation with a Tobit model and challenges in Cambodia

ABSTRACT

Although arsenic contamination was originally considered an environmental problem, its impact on human beings is now widely deemed a serious health issue. This study examines whether international communities perceive arsenic as a health issue and discusses current challenges regarding arsenic contamination in Cambodia. Using comprehensive panel data covering 122 health aid recipient countries across three-plus decades (1990–2020), the first part of our study finds that the recipients’ health needs caused by major incidences (HIV, tuberculosis, and malaria) are all important factors determining health aid. However, health aid is conditionally effective on countries with high arsenic contamination when they have a better governance level. In this context, we conduct a case study focusing on Cambodia, where arsenic contamination is quite serious but given less attention than in other countries. In Cambodia, both inter-governmental coordination and coordination among international organizations are weak, and no long-term blueprint for arsenic mitigation has yet been developed. To fundamentally and sustainably address this problem, we suggest that Cambodia establish a control tower incorporating government, research institutes, and international agencies.

KEYWORDS:

• Health aid; DAH
• IHME
• panel data
• Arsenic
• Cambodia

1. Introduction

Development assistance to the health sector has soared over the past two decades. Ravishankar et al. (2009) found that the amount of such aid rose from $5.6 billion USD in 1990 to $21.8 billion in 2007, and that it has skyrocketed since 2002 because of increased funds from bilateral donor governments (in particular, the United States) and private funding, including philanthropic donations and corporate contributions. Meanwhile, Piva and Dodd (2009) showed that health initiatives have become increasingly prioritized in international development over the past decade. Since the COVID-19 pandemic, in particular, official development assistance to the health sector has scaled up considerably, from approximately $16.22 billion in 2019 to $31.86 billion in 2021 (constant price) (Ahmad and Carey 2021; OECD 2023).

Three of the eight Millennium Development Goals (MDGs) address health issues – Goal 4 (reduce child mortality), Goal 5 (improve maternal health), and Goal 6 (combat HIV/AID, malaria, and other diseases) These issues are more comprehensively and specifically discussed in in Goals 3 (Good Health and Well-being), 6 (Clean Water and Sanitation), 14 (Life Below Water), and 15 (Life on Land) of the Sustainable Development Goals (SDGs). The World Health Organization (WHO 2005) pointed out that ‘Health is the heart of the MDGs,’ thus recognizing its importance in reducing poverty and achieving human well-being, as well as emphasizing further disbursement.

While many studies – from McKinlay and Little (1977), Burnside and Dollar (2000), and Berthélemy and Tichit (2004) to Kim and Oh (2012) and Oh and Kim (2015) – have empirically examined the determinants and effectiveness of aid in general, studies examining health-specific aid remain relatively scarce (Lane and Glassman 2007; Lim et al. 2012; Ravishankar et al. 2009). Lane and Glassman (2007) found that health aid improvements in terms of consolidation and increased attention to efficiency generally meet recipients’ needs. Lim et al. (2012) measured risk factors for different disease burdens and found that those risks differed by periods, transitioning from communicable diseases in children to non-communicable diseases in adults. In another vein, Ravishankar et al. (2009) examined 1990–2007 Development Assistance to Health (DAH) from the Institute for Health Metrics and Evaluation (IHME) and found that DAH quadrupled during that period (from $5.6 billion to $21.8 billion), with the proportion varying by organization and the increase boosting funds to improve global health – addressing HIV/AIDS (most notably), tuberculosis, and malaria.

After reviewing past literature, we employ DAH (Ravishankar et al. 2009) as a dependent variable and empirically examine its determinants, adopting the same approach as Lane and Glassman (2007) but using a comprehensive dataset that includes more variables and up-to-date time periods. In particular, we pay special attention to arsenic as a determining factor, which previous studies have not done despite it being linked directly to the Sustainable Development Goal 6.1 – that is, universal and equitable access to safe and affordable drinking water by 2030 (UNICEF 2019). In fact, although arsenic itself was originally an environmental problem because it contaminates drinking water, it is now a major health issue to which over 140 million people in 70 countries are exposed (WHO 2023). Indeed, arsenic is highly toxic – it can cause skin lesions, cancer, and developmental defects in children. Moreover, it is invisible, odorless, and tasteless, which exacerbates the difficulty of raising public awareness¹ regarding the danger of this deadly poison.

Against this backdrop, we develop hypotheses to investigate various determinants of health aid. While Lane and Glassman (2007) selected income, population, and health expenditures as main explanatory variables, we go further, adding recipients’ health needs stemming from major incidences (HIV, tuberculosis, malaria, as well as arsenic contamination), and governance² levels and employing a dataset that includes data up until 2020, based on the IHME’s DAH’s availability.

This study focuses on Cambodia, a country that has received little academic attention (Petrusevski et al. 2008) even though the well water in certain areas of the country have some of the world’s highest arsenic levels (Jones and Khaira 2014). The risk is relatively higher for residents in Kandal province than in Kampong Cham and others. In examining the Cambodian case, we give particular attention to coordination among the country’s different stakeholders, including government ministries, bilateral development agencies, international organizations, non-governmental organizations (NGOs), and universities (e.g. Gibson et al. 2016), closely examining the country’s Arsenic Inter-Ministerial Sub Committee (AISC).

This paper is organized as follows. Section two explains the research design and presents the hypotheses. Section three describes the study’s data and methodology. Section four summarizes the findings and discusses key results. Section five addresses the Cambodian case. Finally, section six concludes the study by discussing the implications of our findings.

2. Research design and hypotheses

To examine determinants of health aid, we test the following hypotheses. We derive Hypothesis 1 from Lane and Glassman (2007), who set income, population, and health expenditures as main explanatory variables, and Hypothesis 2 from Alesina and Dollar (2000), who focused on aid effectiveness. We go further than both of these previous studies, adding factors to be analyzed, including disease incidence, governance, and – crucially – arsenic contamination.

Hypothesis 1: The most important determinant of health aid (DAH) is the health need of recipient countries; specifically, because countries in which HIV, tuberculosis (TUB), and malaria are prevalent receive more funding, these three health needs along with life expectancy (a natural indicator of a country’s health status) are key determinants of health aid.

Hypothesis 2: Since governance in recipient countries is an important determining factor, countries with stronger governance receive more DAH funding.

Hypothesis 3: When the international community regards arsenic as a health issue rather than an environmental one, countries with serious arsenic contamination receive more health aid (DAH), depending on their governance level.

Fundamentally, Hypothesis 1 posits that health aid volumes (measured by IHME’s DAH) are proportional to countries’ needs (i.e. more funding for HIV, TUB, and malaria in countries where the corresponding diseases are prevalent). Therefore, we expect all these factors to have positive signs. In addition to sector-specific disbursement, we consider health aid total disbursement to test whether it is positively associated with the health needs. Here, we include life expectancy and expect its sign to be negative.

We derive Hypothesis 2 from studies on aid efficacy, which have shown associations between growth in aid and recipients’ governance levels; countries with good governance have both the capacities and capabilities to link aid amounts and growth (Burnside and Dollar 2000). Measuring governance numerically can be challenging. This study employs three indicators from the following sources – the Country Policy Institutional Assessment (CPIA), produced solely by World Governance Indicator (WGI), jointly produced by the World Bank and Brookings Institution. The former ranges from 1 (low) to 6 (high) based on 16 criteria in four clusters (economic management, structural policies, policies for social inclusion and equity, and public sector management and institutions), out of which this study used two criteria that are directly relevant to governance (property rights and rule-based governance, and transparency, accountability, and corruption). The WGI ranges from −2.5 to 2.5, capturing perceptions of the quality of public and civil services as well as the quality of policy formulation and implementation. We used an unweighted average of these three, based on which we hypothesize that donor countries tend to provide more aid to countries with good governance, making the expected sign of governance positive.

Hypothesis 3 is the primary focus of this study; in it, we postulate that countries with high arsenic contamination receive more funding. Our analysis interacts this dummy variable with governance, the main variable in Hypothesis 2, to check whether it magnifies the effects on the determinants of DAH. The interaction term could partially differentiate the dependent variable (DAH) with respect to governance level, thereby revealing its marginal effect. Thus, it should demonstrate whether the arsenic binary dummy makes any difference in the governance-aid nexus.

3. Data and methodology

We employ panel data, covering 122 countries whose health sectors received development assistance over the three-plus decades between 1990 and 2020.³ As dependent variables (denoted as DAH), we mainly use the total amount of DAH each country received (thousand constant 2021 US dollars), collected by the Institute for Health Metrics and Evaluation (IHME). In addition, we use specific DAH amounts to fight HIV, tuberculosis, and malaria (thousand constant 2021 US dollars), all of which we derive from the IHME, as well.

We derive all our independent variables from the World Bank’s World Development Indicator (WDI). The first, a group of variables, denoted as Health in Equations (1) and (2), refers to the incidence of major diseases – HIV (per 1000 uninfected population), tuberculosis (per 100,000 people), and malaria (per 1000 population at risk), followed by life expectancy at birth (years). Next, we adopt the arsenic dummy, coded as 1 for countries that have reported substantial arsenic contamination in their ground water and 0 otherwise⁴ (Shaji et al. 2021; WHO 2023).

The remaining explanatory variables concern socio-economic status; we consider GDP per capita (denoted as PGDP, constant 2017 international dollar), population (denoted as POP, the number of people), and health expenditures per capita (denoted as EXP, current international dollar). Lastly, we use governance, denoted as GOV – an unweighted average of three criteria, two from CPIA and one from WGI, introduced in the previous section – in the second estimation, interacting it with the arsenic dummy. This leads to the following two equations, Equation (1) for the first estimation and (2) for the second. (1) $DA{H}_{ij}={\beta }_0+{\beta }_{1}Healt{h}_{ij}+{\beta }_{2}Arseni{c}_i+{\beta }_{3}PGD{P}_{ij}+{\beta }_{4}PGD{P}_{ij}^2+{\beta }_{5}PO{P}_{ij}+{\beta }_{4}PO{P}_{ij}^2+{\beta }_{5}EX{P}_{ij}+{\epsilon }_{ij}$(1) (2) $DA{H}_{ij}={\beta }_0+{\beta }_{1}Healt{h}_{ij}+{\beta }_{2}Arseni{c}_i+{\beta }_{3}Go{v}_{ij}+{\beta }_{4}Arseni{c}_i\ast Go{v}_{ij}+{\beta }_{5}PGD{P}_{ij}+{\beta }_{6}PGD{P}_{ij}^2+{\beta }_{7}PO{P}_{ij}+{\beta }_{8}PO{P}_{ij}^2+{\beta }_{9}EX{P}_{ij}+{\epsilon }_{ij}$(2)

Except for the binary dummy for arsenic and (the unweighted average of) the categorical variable for governance, we log-transform all the explanatory variables to consider different units. Table 1 provides more detailed explanations of the variables and their summary statistics. We lag all the variables for one year to avoid any potential endogeneity problems.

Table 1. Summary statistics for variables and grouping.

Variable	Unit	Obs	Mean	Std. Dev.	Min	Max
Dependent
TOTAL_DAH	1000 USD (Constant 2021)	3782	106,285	196,760	0	2,455,335
HIV_DAH	1000 USD (Constant 2021)	3782	18,938	63,576	0	669,093
TUB_DAH	1000 USD (Constant 2021)	3782	3674	11,203	0	213,480
Malaria_DAH	1000 USD (Constant 2021)	3782	6432	20,787	0	315,122
Independent
HIV_Incidence	Per 1000 people	2976	1.386	3.0	0.01	25.77
TUB_Incidence	Per 100,000 people	2541	187.1	202.1	0	1270
MAL_Incidence	Per 1000 people	1993	113.1	156.7	0	723.2
Life Expectancy	Years (at birth)	3782	64.26	8.847	14	80
Arsenic	0 or 1 (binary dummy)	2991	3.82	1.693	1	7
Governance	1 to 6; −2.5 to 2.5	1142	4.947	1.675	0.261	9.004
GDP per capita	USD, PPP, Constant 2017	3478	6606	436	436.3	33,345
Population	People	3782	4256K	9K	9182	1411M
Health Exp	USD, PPP, Current	2453	377.5	373.3	6.717	3176
Dependent variables	Independent variables
	Health-related	Socio-economic	Arsenic dummy
DAH ln_TOTAL ln_HIV ln_TUB ln_Malaria	Disease- Incidence ln_hiv1 ln_tub1 ln_malaria1 ln_lifeex1	ln_PGDP1 ln_PGDP1^2 ln_POP1 ln_POP1^2 governance1 ln_EXP1	Afghanistan, Algeria, Armenia, Azerbaijan, Bangladesh, Bolivia, Botswana, Burkina Faso, Cambodia, Cameroon, Costa Rica, China, Cuba, Dominica, Ecuador, El Salvador, Egypt, Ethiopia, Ghana, Honduras, India, Indonesia, Jamaica, Kenya, Kyrgyzstan, Laos, Libya, Malawi, Malaysia, Mexico, Mongolia, Morocco, Myanmar, Namibia, Nepal, Nicaragua, Niger, Nigeria, Pakistan, Papua New Guinea, Paraguay, Philippines, Peru, South Africa, Sri Lanka, Tajikistan, Tanzania, Thailand, Togo, Turkmenistan, Uganda, Uzbekistan, Vietnam, Zambia, Zimbabwe
Interaction Variable: governance * arsenic

Notes: DAH: Development Assistance to Health (from IHME); TUB: Tuberculosis; Lifeex: Life Expectancy at birth; PGDP: Per capita GDP; POP: Population; Exp (of Health Exp): Health expenditure; Arsenic: dummy variable that codes 1 for countries that have reported substantial arsenic contamination in their ground water and 0 otherwise (Shaji et al. 2021). All the variables except governance and arsenic are log-transformed and time-variant independent variables are one year lagged.

Source: Dependent Variable: Development Assistance to Health, Institute for Health Metrics Evaluation. Independent Variable: World Development Indicator, World Bank (2023).

Regarding methodology, this study uses a Tobit estimation because several dependent variables include zero values. Theoretically, a Tobit model provides the censored result, coding a dependent variable of zero as the left-censored lower limit – a practice that is frequently used in other empirical studies of foreign aid, which naturally contain many zero values in each observation (Alesina and Dollar 2000; Berthélemy and Tichit 2004; Kim and Oh 2012; Li, Zeng, and Fan 2008).⁵

4. Results

Table 2 shows results for Hypothesis 1 and part of Hypothesis 3, based on Equation (1). The first five columns use total amount of DAH as a dependent variable (denoted as TOTAL in Table 2). Three of them contain only one health-incidence variable in each analysis (HIV, tuberculosis, and malaria), followed by life expectancy in the fourth column, and finally, all the incidences together (column five). For the remainder, we use specific DAHs to HIV, tuberculosis, and malaria as dependent variables (denoted as HIV, TUB, and MAL in the table) to test the corresponding disease risk (e.g. HIV-specific DAH to HIV incidence, etc.)

Table 2. Determinants of DAH (Development Assistance to Health): Tobit model.

	TOTAL	TOTAL	TOTAL	TOTAL	TOTAL	HIV	TUB	MAL
ln_hiv1	0.207***				0.164***	0.724***
	(0.013)				(0.019)	(0.029)
ln_tub1		0.184***			0.144***		0.869***
		(0.019)			(0.029)		(0.064)
ln_malaria1			0.001***		0.042***			0.479***
			(0.002)		(0.000)			(0.028)
ln_lifeex1				−0.034***
				(0.004)
arsenic	0.170***	0.116*	0.156**	0.199***	0.127**
	(0.064)	(0.608)	(0.065)	(0.061)	(0.064)
ln_pgdp1	4.844***	3.214***	4.046***	3.991***	5.670***	7.947***	10.028***	9.701***
	(0.453)	(0.441)	(0.494)	(0.442)	(0.499)	(0.994)	(1.439)	(1.550)
ln_pgdp1^2	−0.350***	−0.250***	−0.304***	−0.294***	−0.409***	−0.547***	−0.714***	−0.737***
	(0.027)	(0.026)	(0.030)	(0.026)	(0.303)	(0.059)	(0.086)	(0.098)
ln_pop1	−0.017	1.017***	0.580***	0.945***	0.079	1.225***	2.739***	−0.054
	(0.181)	(0.095)	(0.160)	(0.096)	(0.177)	(0.402)	(0.327)	(0.500)
ln_pop1^2	0.018***	−0.015***	−0.001	−0.012***	0.015***	−0.009	−0.055***	−0.026*
	(0.006)	(0.003)	(0.005)	(0.003)	(0.005)	(0.012)	(0.010)	(0.015)
health_exp1	0.582***	0.516***	0.602***	0.577***	0.811***	0.832***	1.590***	1.592***
	(0.053)	(0.048)	(0.056)	(0.050)	(0.057)	(0.116)	(0.161)	(0.178)
Constant	−11.885***	−14.159***	−13.393***	−14.426***	−17.660***	−40.554***	−69.846***	−38.189***
	(2.546)	(2.082)	(2.557)	(2.083)	(2.578)	(5.658)	(6.923)	(8.066)
Obs	1794	2244	1756	2251	1423	1794	2244	1590
L-censored	0	0	0	0	0	30	363	142
R square	0.276	0.291	0.240	0.291	0.305	0.143	0.110	0.109

Notes: Obs: The number of observations; L-censored: Left Censored or zero as a lower limit; Standard errors in brackets. DAH: Development Assistance to Health (from IHME); TUB: Tuberculosis; MAL: Malaria; Lifeex: Life Expectancy at birth; PGDP: Per capita GDP; POP: Population; Health Exp: Health expenditure; Arsenic: dummy variable that codes 1 for countries that have reported substantial arsenic contamination in their ground water and 0 otherwise (Shaji et al. 2021). All the variables except governance and arsenic are log-transformed and time-variant independent variables are one year lagged.

Source: Dependent Variable: Development Assistance to Health, Institute for Health Metrics Evaluation. Independent Variable: World Development Indicator, World Bank (2023).

***p < 0.01.

**p < 0.05.

*p < 0.1.

In general, we find that the health status variables consistently have the expected signs, regardless of the dependent variables (total DAH or specific ones). That is, the coefficients for HIV, tuberculosis, and malaria are all significantly positive, and those for life expectancy are significantly negative, confirming Hypothesis 1. The fact that the signs of the arsenic dummy variables are positive also confirms Hypothesis 3; countries with major arsenic contamination receive, on average, 1.12–1.22⁶ times more funding than others; although this difference may not appear dramatic, it is statistically significant across all estimations.

Following the approach of several international development studies (Alesina and Dollar 2000; Berthélemy and Tichit 2004; Kim and Oh 2012), we use both linear and quadratic terms for GDP per capita and population – a more advanced approach than that of Lane and Glassman (2007), who only used linear terms. Across all estimations, the signs for GDP per capita are all significantly positive for the linear term and negative for the quadratic term. This implies that a country’s GDP per capita positively impacts health aid, albeit at a diminishing rate. We find the same pattern for the population variable, although the results are neither as consistent nor as significant as those for GDP per capita. Lastly, consistent with Lane and Glassman (2007), our analyses shows positive signs for health expenditures in all columns.

Table 3 shows results related to Hypothesis 2, which concerns the effects of arsenic incidence on the aid countries receive. With total DAH as the only dependent variable, we conduct five Tobit regression analyses, corresponding to the first five columns in Table 2. The overall results are essentially the same as the previous outcomes, generating positive coefficients for HIV, tuberculosis, malaria, arsenic, and health expenditures, negative coefficients for life expectancy, and consistent results for GDP per capita and population.

Table 3. Determinants of DAH: extended Tobit model.

	TOTAL	TOTAL	TOTAL	TOTAL	TOTAL
ln_hiv1	0.149***				0.115***
	(0.014)				(0.021)
ln_tub1		0.099***			0.056
		(0.022)			(0.038)
ln_malaria1			0.084***		0.047***
			(0.009)		(0.011)
ln_lifeex1				−0.014***
				(0.004)
arsenic	1.291***	0.425**	0.844***	0.544***	1.172***
	(0.182)	(0.170)	(0.173)	(0.175)	(0.182)
governance	0.135***	0.134***	0.148***	0.135***	0.142***
	(0.015)	(0.016)	(0.015)	(0.016)	(0.015)
arsenic x	−0.272***	−0.121***	−0.213***	−0.139***	−0.255***
governance	(0.035)	(0.031)	(0.034)	(0.032)	(0.035)
ln_pgdp1	3.418***	3.317***	3.529***	3.749***	4.196***
	(0.619)	(0.523)	(0.659)	(0.516)	(0.664)
ln_pgdp1^2	−0.253***	−0.254***	−0.268***	−0.280***	−0.304***
	(0.039)	(0.032)	(0.042)	(0.032)	(0.042)
ln_pop1	−0.030	0.691***	0.388**	0.689***	0.079
	(0.178)	(0.106)	(0.158)	(0.107)	(0.180)
ln_pop1^2	0.022***	−0.001	−0.009*	0.001	0.019***
	(0.006)	(0.004)	(0.005)	(0.004)	(0.006)
health_exp1	0.422***	0.429***	0.668***	0.471***	0.574***
	(0.055)	(0.051)	(0.056)	(0.052)	(0.060)
Constant	−7.251**	−12.710***	−12.173***	−13.367***	−12.230***
	(2.952)	(2.320)	(2.985)	(2.324)	(3.100)
Obs	848	1081	860	1083	766
L-censored	0	0	0	0	0
R square	0.483	0.502	0.504	0.501	0.515

Notes: Obs: The number of observations; L-censored: Left Censored or zero as a lower limit; Standard errors in brackets. DAH: Development Assistance to Health (from IHME); TUB: Tuberculosis; Lifeex: Life Expectancy at birth; PGDP: Per capita GDP; POP: Population; Health Exp: Health expenditure; Arsenic: dummy variable that codes 1 for countries that have reported substantial arsenic contamination in their ground water and 0 otherwise (Shaji et al. 2021). All the variables except governance and arsenic are log-transformed and time-variant independent variables are one year lagged.

***p < 0.01.

**p < 0.05.

*p < 0.1.

What differentiates this table from the preceding table is the inclusion of a new variable, governance (${\beta }_3$ in Equation [2]), which we find to be significantly positive (ranging from 0.134 to 0.148). This is an expected result, given that governance is measured based on a mixture of CPIA (1 for low and 6 for high) and WGI (−2.5 for low and 2.5 for high), capturing rule-based governance, transparency, accountability, and quality of public and civil services; in short, we find that countries with better governance tend to secure more funding. Very interestingly, when we interact this variable with the arsenic dummy (${\beta }_4$ in Equation [2]), the coefficients become significantly negative in all cases. The partial derivative with respect to governance level (known as the marginal effect) reveals that when the arsenic dummy is zero, the marginal effect is equal to β₃ (significantly positive), which becomes β₃+ β₄, when the dummy is equal to one (countries with high arsenic contamination). The negative coefficients of ${\beta }_4$ cancel out the positive impact of β₃, and β₃+ β₄ turns negative in all columns (e.g. 0.135–0.272 = −0.137 in the first column) except the second one. In other words, the finding that the level of DAH a country receives is positively associated with its governance level does not apply to countries with high arsenic incidence; for these countries, having good governance is not enough to secure DAH funding and they may need more. In this regard, this study argues that, for the DAH funding to be effectively channeled to the arsenic sector, these countries may need to focus on improving coordination between various stakeholders, including government ministries. This is because arsenic inherently requires an inter-ministerial approach in countries that originally viewed it an environmental problem but now consider its impact on humans a public health issue, for which coordination is often easier said than done. Focusing on Cambodia, the next section discusses the challenges and possible ways forward in such circumstances.

5. The Cambodian case

In the flat and densely populated area around the Mekong River in Cambodia, the arsenic-contaminated groundwater has not been properly treated. This water is often used for drinking water via wells, and even where it is not, it is used for agricultural purposes, such as rice cultivation. As shown in the map below (Figure 1), high concentrations of arsenic have been detected in the groundwater in the Mekong River basin (represented by the shaded areas, which have been denoted ‘At Risk’). This phenomenon can also be observed in Table 4, which demonstrates that in the Kandal and Kampong Cham areas, the proportion of wells with arsenic levels of 50μg/L or more is quite high (Ratha et al. 2017), with Kandal being more severely affected. In fact, estimates indicate that approximately 2.4 million people have been exposed to such groundwater (Shaji et al. 2021). Arsenic-contaminated food and water pose a greater threat than manganese, cadmium, and lead, and past studies have strongly emphasized the negative impacts of arsenic on public and community health (Murphy et al. 2021; Phan et al. 2016; Pravalprukskul, Aung, and Wichelns 2018; Ratha et al. 2017), highlighting the linkages between arsenic toxicity and the immune system deterioration (Bellamri et al. 2018).

Figure 1. Arsenic risk areas in Cambodia. Source: Ratha et al. (2017).

Table 4. Testing results in affected provinces.

		Arsenic Concentration (ppb)
Province	Well tested (number)	As (> 50)	%	As (0–50)	%	As (0)	%
Kandal	15,143	5325	35	5789	38	4029	27
Kampong Cham	7455	2477	33	4318	59	660	8
Prey Veng	5641	1055	19	1887	33	2699	48
Kampong Thom	1346	40	3	799	59	507	38
Kratie	734	110	15	199	27	417	58
Kampong Chnang	294	19	6	131	44	144	50
Peri-urban Phnom Penh	226	22	10	91	40	113	50

Source: Ratha et al. (2017).

Recognizing this risk, Cambodia has implemented national-level programs to address arsenic issues. For example, the National Action Plan on Rural Water Supply and Sanitation Phase 2 (2019–2023) aims to extend access to clean water to 90 percent of the country (including rural areas) by 2023 (Phnom Penh Post 2022). Additionally, The National Strategy for Rural Water Supply, Sanitation and Hygiene (2011–2025) has set the ambitious goal of achieving 100 percent coverage of rural sanitation services by 2025 (World Bank 2020).

In 2002, The Cambodian government established the Arsenic Inter-ministerial Sub Committee (AISC), which comprises five relevant ministries: the Ministry of Rural Development (MRD), the Ministry of Water Resource and Meteorology (MOWRAM), the Ministry of Health (MoH), the Ministry of Mines and Energy (MIME), and the Ministry of Environment (MoE). As shown in Figure 2, the MRD chairs the AISC alongside two major departments – the Department of Rural Water Supply (DRWS) and the Department of Rural Health Care (DRHC) (MRD 2011).

Figure 2. Structure of the Arsenic Inter-Ministerial Sub Committee (AISC). Source: (MRD 2011).

The fact that the committee includes a variety of ministries bespeaks the widespread recognition that effective arsenic mitigation inherently requires an inter-disciplinary approach involving careful coordination among these ministries. Unfortunately, that is not the case for the AISC; the MRD, the chairing ministry, does not have a strong drive to serve as a control tower, given that the main two departments in the MRD handling the committee lack budget, personnel, and decision-making power.⁷

From a broader perspective, coordination between the Cambodian government and the international community (international organizations, bilateral development agencies, NGOs, etc.) and among these entities is arguably even more challenging. Indeed, the fact that the latest MDGs and SDGs emphasize their partnership – namely, harmonization or coordination among aid donors – reflects this hardship. In Cambodia, incoherence and hierarchy exacerbated by fragmentation pose fundamental challenges to inter-agency collaboration, hampering so-called Delivery as One (DaO) (Melle and Cappellaro 2018).

Moreover, although several projects have been implemented in Cambodia, the country has yet to reach the level of action necessary to address the arsenic problem, which is mostly concentrated in South Asian regions with Bangladesh as the hub (Hossain 2006; Murphy et al. 2021). This situation is truly an obstacle for Cambodia in terms of coordinating with and achieving solidarity from international organizations regarding the urgency of the arsenic problem.⁸

6. Conclusion and discussions

To sustain its arsenic mitigation funding, Cambodia must identify a best-fit model for the specific arsenic contamination case it faces. Rouhi, Gorji, and Maleki (2019) explored eight coordination models that could be used to manage natural disasters, and Cambodia could adopt such a model to address its dangerous arsenic contamination.

First, the MRD-focused AISC must be reformed. Since arsenic contamination is a major environmental problem that significantly impacts people’s health, either the Ministry of Environment or the Ministry of Health should replace the MRD as the control tower. Meanwhile, the Ministry of Education should spearhead efforts to raise public awareness about the arsenic problem, while the Ministry of the Interior – generally tasked with collaborating with provinces – should facilitate communication among each local provinces. The establishment of a cooperative model with health- and/or environment-related international organizations and university research institutions is also crucial. Although several laboratories have the capacity to conduct arsenic studies (Research Development International, Institute of Technology of Cambodia, Royal University of Phnom Penh, University of Health Sciences, International University, Ministry of Commerce, Ministry of Industry, Science, Technology & Innovation and Ministry of Environment), most of them lack stable electrical service, face budget constraints, and do not cooperate among themselves. Most importantly, given the top-down nature of the country’s decision-making process, it is of the utmost importance that the Prime Minister’s office makes tackling this problem a focal point – to use a game-theory term – around which (among several possible options) coordination becomes a self-fulfilling belief (Basu 2018).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea [NRF-2022S1A5A2A01045555].

Notes

1 Awareness, locally and globally, regarding the danger of arsenic is still quite low. Without targeted public awareness campaigns, arsenic remains an abstract and remote concept for many people, particularly for those with less education. The WHO’s threshold for defining arsenic contamination in ground water is 10 ug/L (micrograms per liter) or 10 ppb, but the threshold is set much higher in Cambodia at 50 ug/L (WHO 2023).

2 This variable plays a key role in this paper, serving as a bridge between the first part of the empirical analyses with the second part of the Cambodian case.

3 The DAH data covers 173 developing countries and the WDI covers over 200. However, the data of only 122 countries is included in both sources over this study’s entire time span. Additionally, the DAH dataset only includes data through 2020.

4 Afghanistan, Algeria, Armenia, Azerbaijan, Bangladesh, Bolivia, Botswana, Burkina Faso, Cambodia, Cameroon, Costa Rica, China, Cuba, Dominica, Ecuador, El Salvador, Egypt, Ethiopia, Ghana, Honduras, India, Indonesia, Jamaica, Kenya, Kyrgyzstan, Laos, Libya, Malawi, Malaysia, Mexico, Mongolia, Morocco, Myanmar, Namibia, Nepal, Nicaragua, Niger, Nigeria, Pakistan, Papua New Guinea, Paraguay, Philippines, Peru, South Africa, Sri Lanka, Tajikistan, Tanzania, Thailand, Togo, Turkmenistan, Uganda, Uzbekistan, Vietnam, and Zambia. A continuous variable would have been more desirable, but data are unavailable.

5 Given the relatively lower number of zero values, we conduct panel fixed effect analyses as a robustness check. These analyses generate results that are the same as those shown in Table 2, with a lower significance level and an omitted finding for the arsenic dummy variable, due to its time-invariant nature. They can be provided upon request.

6 Given that the coefficients vary from 0.116 to 0.199, Exp (0.116) = 1.12 and Exp(0.199) = 1.22.

7 This is based on interviews with the Director and his team in the Department of Rural Water Supply of the Ministry of Rural Development. (11 November 2022) in Phnom Penh, Cambodia. The director and his department oversee the AISC.

8 After visiting ADB, WFP, and UNESCO in July 2022, the authors confirmed that no current projects tackle arsenic problems. In the ADB Cambodia office, actions to address arsenic problems have been stalled since 2009. This is the case even though the Country Director is fully cognizant of arsenic issues. While experts at WFP claim that problems such as food poisoning may occur when arsenic is present in rice cultivation, none of their projects seek to address the issue. Additionally, experts at UNESCO do not conduct any educational programs to raise public awareness about the arsenic problem.